We are, at this very moment, building artificial human beings.
Not, of course, by assembling pieces of dead flesh and waiting for thunderstorms, as Dr. Frankenstein did. Instead, researchers at IBM and EPFL are trying to implement a simulation of an entire human brain, running on one of the world's most powerful computers. Instead of studying visible manifestations of cognitive processes, as traditional AI researchers do, Henry Markram and his team work on a scale smaller than cells, analyzing and reproducing chemical interactions inside and in-between individual neurons. Experiences simulating a rat's neo-cortical column are encouraging, and the next target is the 100-billion-neurons human brain.
Our understanding of consciousness is still very poor, but at least one school of thought, known as Functionalism (philosophy of mind), argues that the human experience of consciousness is the result not of a metaphysical soul, nor of any yet-to-be-discovered physical property of gray and white matter, but only a by-product of the special kind of information processing that the brain does. In other words, if a subroutine running on a computer executes an algorithm that mimics the information processing of a human brain, that subroutine becomes conscious.
Consciousness might mean very little without sensory experiences, but there doesn't seem to be a reason those couldn't be provided to the virtual brain. Retinas are nothing but collections of specialized neurons, which can be simulated like any others, with the pattern of light striking them provided as an input to the simulation. Star Trek fans might suggest Data's synthetic eyeballs as the best source for such inputs, but an even more exciting idea is to get them from another subroutine simulating the environment.
A powerful-enough computer, such as we're likely to get in two or three decades, could in theory simulate a realistic environment inhabited by fully-thinking simulated human beings. Not only would this provide an intense thrill to researchers offered the chance to "play god", it would have serious practical applications. Non-playing characters in video games would certainly act more realistically if their actions were guided not by simplistic behavioral scripts but by genuinely human thought processes. A more serious possibility is turning economics into a genuine experimental science: one could build a virtual country with similar economic parameters as the current United States, then enact Herman Cain's 9-9-9 tax reform plan, and watch what happens.
As useful as such technology might prove, it does raise serious ethical questions. A simulated human being would act like a flesh-and-blood human being because he could feel everything a human being does, including fear, pain, and despair. Indeed, by some definitions of the term, such a simulation would be genuinely human.
Is it ethical to create short-lived, shell-shocked, terrified virtual soldiers just so you have more realistic opponents in Call of Duty? Is it moral to create a few thousand starving children and desperate parents so you can check whether or not your fiscal ideas are sound?
I'm sure many people think these questions are pointless because the entire premise seems like a pipe-dream. Maybe they're right. Maybe functionalists are wrong, and a simulation can never be sentient. It is, however, entirely possible that they are right. Dr. Markram says his model of the brain will be ready and functioning by 2020. Maybe, just maybe, these bizarre ethical questions are not at all ridiculous, and they'll need answering, very, very soon. Long before 2061.
Luc Steels and Jack Tuszynski artificial intelligence experts will both be at TEDx Brussels November 22